Abstract: A distributed data processing system may include an interface that receives a data processing request from a requesting entity, a processing server to provide access to local data processing applications, a shadow processing server to provide access to remote data processing applications, and an application server to fulfill the received data processing request by selectively accessing local and remote data processing applications transparently to the requesting entity. Access to data may be facilitated by providing heterogeneous data sources with software wrappers that provide an object representation of the data source, providing outputs of software wrappers to a first accumulator that aggregates data to generate a first aggregate data representation, and using a second accumulator to generate a second aggregate data representation based on the first aggregate data representation from the first accumulator. The software wrappers may hide details (e.g., format, location) of the data source.